Machine learning tool outperforms clinicians in differentiating muscular conditions

Researchers have developed an algorithm to assist the diagnosis of muscular disorders based on stained pathology images, using deep convolutional neural networks (CNNs).

The algorithm was found to outperform clinicians in accurately differentiating between two groups of diseases: idiopathic inflammatory myopathies (IIMs), and non-myositis and neurogenic diseases.

The researchers from IBM Japan and the National Center of Neurology and Psychiatry, in Tokyo, Japan, developed training data-sets and test data-sets based on a total of 4,041 hematoxylin and eosin (H&E)-stained pathology images from 1,400 slides.

The aim was to develop a 2-step algorithm. The first step would differentiate between the following two groups: